BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:BSU Seminar: &quot\;Generating crossmodal gene expression from can
 cer histopathology improves multimodal AI predictions&quot\; - Samiran Dey
 \, Indian Association for the Cultivation of Science\, Kolkota
DTSTART:20260414T133000Z
DTEND:20260414T143000Z
UID:TALK246133@talks.cam.ac.uk
CONTACT:Alison Quenault
DESCRIPTION:Transcriptomic profiling provides rich molecular insights for 
 cancer diagnosis and prognosis\, but its high cost limits routine clinical
  use\, where histopathology remains the primary diagnostic modality. Recen
 t advances in artificial intelligence suggest that molecular information c
 an be inferred directly from digital pathology images. This talk discusses
  a generative multimodal framework that synthesizes transcriptomic feature
 s from whole-slide histopathology images and incorporates them to improve 
 cancer\n grading and survival risk prediction across multiple cancer cohor
 ts. The approach achieves performance comparable to models using real tran
 scriptomic data while relying only on routinely available histopathology i
 mages. To address reliability in generative and predictive models\, a conf
 ormal prediction–based framework is further used to quantify and calibra
 te uncertainty\, providing coverage guarantees and improving fairness acro
 ss gene categories and demographic subgroups. These results highlight the 
 potential\n of generative multimodal learning and calibrated uncertainty t
 o enable accurate\, interpretable\, and reliable AI systems for cancer dia
 gnosis and prognosis.
LOCATION:Large Downstairs Teaching Room\, East Forvie Building\, Forvie Si
 te Robinson Way Cambridge CB2 0SR.
END:VEVENT
END:VCALENDAR